ci: Build Firefox with current neqo

[larseggert]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.24%. Comparing base (dcc88e3) to head (9ce0043).
Report is 9 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1834      +/-   ##
==========================================
- Coverage   93.25%   93.24%   -0.02%     
==========================================
  Files         110      110              
  Lines       35793    35810      +17     
==========================================
+ Hits        33380    33390      +10     
- Misses       2413     2420       +7     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[larseggert]

So it's building (yay!) but given the time it takes to do so, not sure it's productive to enable this for each PR push.

[github-actions]

Benchmark results

Performance differences relative to ed19eb2.

• drain a timer quickly time: [308.94 ns 317.59 ns 325.55 ns]
  change: [-2.1602% -0.0800% +2.3222%] (p = 0.94 > 0.05)
  No change in performance detected.

• coalesce_acked_from_zero 1+1 entries
  time: [197.36 ns 197.74 ns 198.15 ns]
  change: [-1.3154% -1.0002% -0.6823%] (p = 0.00 < 0.05)
  Change within noise threshold.

• coalesce_acked_from_zero 3+1 entries
  time: [241.03 ns 241.70 ns 242.39 ns]
  change: [-0.0419% +0.2714% +0.5815%] (p = 0.09 > 0.05)
  No change in performance detected.

• coalesce_acked_from_zero 10+1 entries
  time: [239.89 ns 240.70 ns 241.64 ns]
  change: [-0.9279% -0.4535% -0.0126%] (p = 0.06 > 0.05)
  No change in performance detected.

• coalesce_acked_from_zero 1000+1 entries
  time: [218.45 ns 218.63 ns 218.84 ns]
  change: [+0.5619% +1.0983% +1.6782%] (p = 0.00 < 0.05)
  Change within noise threshold.

• RxStreamOrderer::inbound_frame()
  time: [118.22 ms 118.33 ms 118.43 ms]
  change: [-0.7297% -0.6270% -0.5282%] (p = 0.00 < 0.05)
  Change within noise threshold.

• transfer/Run multiple transfers with varying seeds
  time: [117.87 ms 118.10 ms 118.33 ms]
  thrpt: [33.803 MiB/s 33.870 MiB/s 33.937 MiB/s]
  change:
  time: [-2.1355% -1.8451% -1.5361%] (p = 0.00 < 0.05)
  thrpt: [+1.5601% +1.8798% +2.1821%]
  Change within noise threshold.

• transfer/Run multiple transfers with the same seed
  time: [118.37 ms 118.57 ms 118.78 ms]
  thrpt: [33.676 MiB/s 33.734 MiB/s 33.793 MiB/s]
  change:
  time: [-2.0188% -1.7741% -1.5139%] (p = 0.00 < 0.05)
  thrpt: [+1.5372% +1.8061% +2.0604%]
  Change within noise threshold.

• 1-conn/1-100mb-resp (aka. Download)/client
  time: [1.1445 s 1.1496 s 1.1549 s]
  thrpt: [86.590 MiB/s 86.990 MiB/s 87.374 MiB/s]
  change:
  time: [+4.2308% +4.9359% +5.6574%] (p = 0.00 < 0.05)
  thrpt: [-5.3545% -4.7037% -4.0591%]
  💔 Performance has regressed.

• 1-conn/10_000-parallel-1b-resp (aka. RPS)/client
  time: [426.65 ms 429.03 ms 431.46 ms]
  thrpt: [23.177 Kelem/s 23.308 Kelem/s 23.438 Kelem/s]
  change:
  time: [-0.3219% +0.4127% +1.1465%] (p = 0.28 > 0.05)
  thrpt: [-1.1335% -0.4110% +0.3230%]
  No change in performance detected.

• 1-conn/1-1b-resp (aka. HPS)/client
  time: [49.791 ms 50.014 ms 50.286 ms]
  thrpt: [19.886 elem/s 19.994 elem/s 20.084 elem/s]
  change:
  time: [-1.4047% -0.6553% +0.0840%] (p = 0.09 > 0.05)
  thrpt: [-0.0839% +0.6596% +1.4247%]
  No change in performance detected.

Client/server transfer results

Transfer of 134217728 bytes over loopback.

Client	Server	CC	Pacing	Mean [ms]	Min [ms]	Max [ms]	Relative
msquic	msquic			748.9 ± 219.0	411.6	1003.2	1.00
neqo	msquic	reno	on	1020.3 ± 288.2	764.8	1382.7	1.00
neqo	msquic	reno		1019.8 ± 316.8	795.1	1718.4	1.00
neqo	msquic	cubic	on	1097.7 ± 292.1	767.3	1666.8	1.00
neqo	msquic	cubic		901.8 ± 200.3	764.5	1399.7	1.00
msquic	neqo	reno	on	4527.0 ± 297.1	4264.2	5234.2	1.00
msquic	neqo	reno		4439.0 ± 308.6	4105.0	4973.5	1.00
msquic	neqo	cubic	on	4392.3 ± 169.4	4154.3	4615.0	1.00
msquic	neqo	cubic		4483.1 ± 212.6	4204.9	4884.0	1.00
neqo	neqo	reno	on	3687.8 ± 285.8	3240.5	4018.2	1.00
neqo	neqo	reno		3889.3 ± 373.5	3191.3	4392.0	1.00
neqo	neqo	cubic	on	4088.2 ± 445.4	3010.9	4705.9	1.00
neqo	neqo	cubic		4365.5 ± 331.2	3796.1	4895.4	1.00

⬇️ Download logs

[mxinden]

So it's building (yay!) but given the time it takes to do so, not sure it's productive to enable this for each PR push.

How about treating it as an asynchronous CI step, i.e.:

• Don't make it required.
• Make it post a comment on failure.

Thereby we might not catch a regression before merging, but we will know through the comment posted after merge, and can thus easily revert without having to git bisect to find the regression in the first place.

[KershawChang]

I think we probably want to setup mozconfig file to create debug and release build.

[martinthomson]

LGTM, modulo a few nits.

[github-actions]

Firefox builds for this PR

The following builds are available for testing. Crossed-out builds did not succeed.

• Linux: Debug Release
• macOS: Debug Release
• Windows: Debug Release